RE: [lojban] Another stab at a Record on ce'u

["And Rosta" <a.rosta@dtn.ntl.com>]

pc:
> a.rosta@dtn.ntl.com writes:
>   So can I try to outline what I take to
>   be an explicit version of your scheme?
>
>   1. All empty sumti places within du'u fill with zo'e.
>
> Yes.

Thankfully this is something everybody agrees on, and it guarantees
that however messy ka is, there's an unambiguous way to say what we
want.

>   2. In ka abstractions, the first empty place fills with ce'u and the
>      rest fill with zo'e.
>
> Yes
>
>    Exception (or generalization): where context indisputably demands
>    a ka abstraction expressing an n-adic relation, where the value of
>    n is certain, the first n empty places fill with ce'u and the rest
>    with zo'e.
>
> No.  Other {ce'u} are explicit, unmarked are {zo'e}
> Unless there is an explicit {zo'e}, in which case, all {zo'e} appear and
> {ce'u} are implicit.
>
>
> 3. EITHER (XOR):
>   3a. In a ka abstraction, if an overt ce'u fills the x1 then all
>       following empty places fill with ce'u.
>   XOR:
>   3b. In a ka abstraction, if a ce'u precedes the first empty place
>       then all following empty places fill with ce'u
>
> Wordy but OK: if {ce'u} in the first open (not filled with content sumti)
> space then all blanks are {ce'u}
>
> <A. Empty places and their sequence have to be defined as x1<x2<x3<x4<x5
>    (because I can't see any other way of defining them). This means
>    that {ka broda .... fa ko'a} will gardenpath people, because they'll
>    misparse as {ka ce'u broda}. There's probably no way round this; you
>    have to wait to the end of tbe clause to know where the ce'u and
>    zo'e go. Your only safe bet to avoid gardenpathing is to use overt
>    ce'u within du'u>
> No, defined by space: first available space, etc.  If first occurrent
> expression is {fe S} then there is assumed to be a space unfilled before it
> and so on.  We can (and will if need be) establish fixed order to prevent
> gardenpathing of that sort.

When we get to the stage of making ordinarily flexible word order rigig,
it just doesn't seem worth the candle. The conventions are just more
trouble than they're worth.

> <B. Rules 2-3 raise further problems of specification:
>
>    i.  Does an empty place within a nonempty x1 precede an empty x2?
>
>    ii. If nonempty x3 precedes nonempty x2, does an empty place within
>        x3 precede an empty place within x2?>
> Now that is interesting.  I (nor anyone else that I can find) didn't think of
> the gaps in a {ka} phrase being within subordinate pharses.   I can see that
> that will generate some problems for interpretation in some cases (if the
> internal phrase is intensional, for example)

disambiguation in such cases can be done by putting ce'u in the prenex
of the abstraction it belongs to, and then referring to it anaphorically.
Plus the usual default rule that says that things not in prenexes go
to the prenex of the localmost bridi.

> but the question of how to apply
> the conventions -- other than the fact that it opens the possibility for many
> more that 5 omitted placeholders) can be handled rather easily, whichever
> ways sems to work best.  I tend to favor linear orders, so I would go with
> "yes" to both your questions (except for being unsure what "x3 precedes x2"
> means)

I meant "linearly precedes". "klama fi lo tcadu be vi ce'u fe lo tcadu be vi
ce'u", say.

>
> <C. {ka ce'u ce'u ce'u ce'u ce'u klama} = {la'e zo klama}, so it may
>    be that Rule 3 doesn't have to be relied on that much.>
> While I think that what C intends is correct, the grammar is not quite right,
> for the {ka} phrases is a selbri and the {la'e} is a sumti.

I wish ka and du'u created sumti not selbri, so I'd prefer a sumti version.

> So, they do play
> different roles sometimes.  I think doing lambda reduction with {la'e zo
> klama} would be a serious mistake generally.  But the equation does reduce
> the need for {ka ce'u klama} say.
>
> <D. Nor does Rule 2 have to be relied on that much, because {ka ce'u zo'e
>    zo'e zo'e zo'e klama} = {lo'e klama}. >
> I gather (I have not been following that part of this muddle closely--
> keeping tack of the {ce'u} has been quite enough for now) that you and xod
> have come to this conclusion.

Xorxes, not xod. Once John's elephant is up and running it will be easier
to keep track of this sort of thing.

> I can't think where it comes from (but I will
> look) nor can I imagine why it would be true on the basis of anyhting in the
> Book or the history of Lojban (making allowance for the misgrammar again,
> even).  A typical broda is a broda, not a property (though, admittedly, not
> necessarily an existing broda either).

Well, you can read the relevant discussion, but the essence is that we don't
need gadri for typical things, because we have the brivla fadni to do that
job. The other plausible interpretation is that it's the archetype, and
this seems to be the intension, which is then the same thing as a ce'u
abstraction.

> <     ko'a ce ko'e simxu loi ka (ce'u) (ce'u) prami
>
>    could be rendered by something like
>
>      ko'a ce ko'e simxu lo'e prami be (tu'a) ce'u>
>
> While I agree that a {ka} may make more sense than a {nu} in simxu-2 (the
> consequences need looking at) I can't make much sense out of either of these
> sentences.  What would a mass of properties be, since there is only one
> property here, le ka ce'u prami.

If you massify one thing you end up with one thing. The point of massifying
is that it doesn't matter how many rabbits there are; you just treat them
as though they're all the one Mr Rabbit.

> Nor is it clear what that has to do with a
> typical lover of  whatever that all evaluates to (not much at a glance).

Nothing.

> I have enough trouble with Lojban, without trying to deal with Andban and
> other Nalgols (I almost miss guaspe!)

Andban = the Lojban of somebody who pays heed to the reasoned discussion
on the list, and is guided by reasoned conclusions arrived at in those
discussions, rather than by vague hunches about what things are supposed
to mean, based on their one-word English glosses in the mahoste.

> <Either:
>
>   I. Revert to my du'u/ka/si'o proposal>
> Aside from being metaphysicaly and grammatically suspect, it is markedly less
> efficient than this one in test cases and wastes a good (maybe someday
> useful) cmavo.

It puts to use an otherwise useless cmavo, and incidentally serves to
render sensical the minimal usage the cmavo has actually had. Other
candidate meanings for si'o can be rendered by other no less effectual
means.

It is markedly more efficient than the horrible mess of rules you're advocating,
both in enabling fewer words to be used, and in making the rules simpler.

But I don't think any of this is worth fighting battles over. I will rest
content with what we've all agreed uncontroversially about du'u, and will
seek a simple way to convert a selbri to a sumti -- li ni'e klama, say,
if that can be understood without klama being turned into a number along
the way.


> <or:
>
>   II. a. Leave ka grungey, i.e. totally reliant on glorking.
>       b. To avoid relying on glorking, use du'u, lo'e and la'e zo.>
> Nah.  The one is imprecise, the other is unintelligible in the context of
> Lojban and as taken to deal with {ka} issues.

You can't get rid of this imprecision without either (i) formulating rules
that are simple enough to be workable but constrain people to a degree
they won't accept, or (ii) formulating rules that are so complicated they'll
leak, and so complicated they'll be ignored. Leave ka for usage to decide.
Anything else, it transpires, is a waste of energy.

--And.